Presentation on theme: "Results Summaries Spelling Correction David Kauchak cs160 Fall 2009 adapted from:"— Presentation transcript:
Results Summaries Spelling Correction David Kauchak cs160 Fall 2009 adapted from:
Administrative Course feedback
2010 For Mustang | Official Site of the Ford Mustang Ford Mustang – Wikipedia, the free encyclopedia en.wikipedia.org/wiki/Ford_Mustang Mustang Motorcycle Products, Inc. Mustang Survival Corporation en.wikipedia.org/wiki/Ford_Mustang
2010 For Mustang | Official Site of the Ford Mustang 2010 Ford Mustang - The official homepage of the Ford Mustang | FordVehicles.com Ford Mustang – Wikipedia, the free encyclopedia The Ford Mustang is an automobile manufactured by the Ford Motor Company. It was initially based on the second generation North American Ford Falcon,... en.wikipedia.org/wiki/Ford_Mustang Mustang Motorcycle Products, Inc. What a Difference Comfort Makes! Mustang is the world's leader in comfortable aftermarket motorcycle seats for Harley-Davidson®, Victory and Metric Cruiser... Mustang Survival Corporation Design, development, and manufacture of marine and aerospace safety and survival wear. Includes detailed product catalog, sizing charts, FAQs,...
IR Display In many domains, we have metadata about the documents we’re retrieving For web pages, we have titles and URLs For other collections, we may have other types of information For academic articles, what information do we have?
Other information Other times, we may not have explicit meta-data, but may still want to provide additional data Web pages don’t provide “snippets”/summaries Even when pages do provide metadata, we may want to ignore this. Why? The search engine may have different goals/motives than the webmasters, e.g. ads
10 Summaries We can generate these ourselves! Lots of summarization techniques Most common (and successful) approach is to extract segments from the documents How might we identify good segments? Text early on in a document First/last sentence in a document, paragraph Text formatting (e.g. ) Document frequency Distribution in document Grammatical correctness User query!
11 Summaries In typical systems, the summary is a subset of the document Simplest heuristic: the first 50 (or so – this can be varied) words of the document More sophisticated: extract from each document a set of “key” sentences Simple NLP heuristics to score each sentence Learning approach based on training data Summary is made up of top-scoring sentences
Segment identification extract features f 1, f 2, …, f n learning approach segment identifier
13 Summaries A static summary of a document is always the same, regardless of the query that hit the doc A dynamic summary is a query-dependent attempt to explain why the document was retrieved for the query at hand
Dynamic summaries Present one or more “windows” within the document that contain several of the query terms “KWIC” snippets: Keyword in Context presentation Generated in conjunction with scoring If query found as a phrase, all or some occurrences of the phrase in the doc If not, document windows that contain multiple query terms The summary itself gives the entire content of the window – all terms, not only the query terms
Dynamic vs. Static What are the benefits and challenges of each approach? Static Create the summaries during indexing Don’t need to store the documents Dynamic Better user experience Makes the summarization process easier Unfortunately, must generate summaries on the fly and so must store documents and retrieve documents for every query!
Generating dynamic summaries If we cache the documents at index time, can find windows in it, cueing from hits found in the positional index E.g., positional index says “the query is a phrase in position 4378” so we go to this position in the cached document and stream out the content Most often, cache only a fixed-size prefix of the doc Note: Cached copy can be outdated
17 Dynamic summaries Producing good dynamic summaries is a tricky optimization problem The real estate for the summary is normally small and fixed Want short item, so show as many KWIC matches as possible, and perhaps other things like title But users really like snippets, even if they complicate IR system design
Alternative results presentations? An active area of HCI research An alternative: copies the idea of Apple’s Cover Flow for search resultshttp://www.searchme.com/ 20
Spell correction How might we utilize spelling correction? Two principal uses Correcting user queries to retrieve “right” answers Correcting documents being indexed
Document correction Especially needed for OCR’ed documents Correction algorithms are tuned for this Can use domain-specific knowledge E.g., OCR can confuse O and D more often than it would confuse O and I (adjacent on the QWERTY keyboard, so more likely interchanged in typing). Web pages and even printed material have typos Often we don’t change the documents but aim to fix the query-document mapping
Query misspellings Our principal focus here E.g., the query Alanis Morisett We can either Retrieve documents indexed by the correct spelling Return several suggested alternative queries with the correct spelling Did you mean … ? Advantages/disadvantages?
Spell correction Two main flavors/approaches: Isolated word Check each word on its own for misspelling Which of these is mispelled? moter from Will not catch typos resulting in correctly spelled words Context-sensitive Look at surrounding words, e.g., I flew form Heathrow to Narita.
Isolated word correction Fundamental premise – there is a lexicon from which the correct spellings come Two basic choices for this A standard lexicon such as Webster’s English Dictionary An “industry-specific” lexicon – hand- maintained The lexicon of the indexed corpus E.g., all words on the web All names, acronyms etc. (Including the mis-spellings) a able about account acid across act addition adjustment advertisement after again against agreement air all almost …
Isolated word correction Given a lexicon and a character sequence Q, return the words in the lexicon closest to Q How might we measure “closest”? Edit distance (Levenshtein distance) Weighted edit distance n-gram overlap Lexicon q 1 q 2 …q m ?
Edit distance Given two strings S 1 and S 2, the minimum number of operations to convert one to the other Operations are typically character-level Insert, Delete, Replace, (Transposition) E.g., the edit distance from dof to dog is 1 from cat to act is ? (with transpose?) from cat to dog is ? Generally found by dynamic programming See for a nice example plus an applet.http://www.merriampark.com/ld.htm
Weighted edit distance As above, but the weight of an operation depends on the character(s) involved Meant to capture OCR or keyboard errors, e.g. m more likely to be mis-typed as n than as q Therefore, replacing m by n is a smaller edit distance than by q This may be formulated as a probability model Requires weight matrix as input Modify dynamic programming to handle weights
Using edit distance We have a function edit that calculates the edit distance between two strings We have a query word We have a lexicon Lexicon q 1 q 2 …q m ? now what?
Using edit distance We have a function edit that calculates the edit distance between two strings We have a query word We have a lexicon Lexicon q 1 q 2 …q m ? We need to reduce the candidate lexicon set Ideas?
Enumerating candidate strings Given query, enumerate all character sequences within a preset (weighted) edit distance (e.g., 2) Intersect this set with the lexicon dogdoa, dob, …, do, og, …, dogs, dogm, …
Character n-grams Just like word n-grams, we can talk about character n-grams A character n-gram is n contiguous characters in a word remote remoteremote re em mo ot te rem emo mot ote remo emot mote unigrams bigrams trigrams4-grams
Character n-gram overlap Enumerate all n-grams in the query string Identify all lexicon terms matching any of the query n-grams Threshold by number of matching n-grams Weight by keyboard layout, etc. What is the trigram overlap between “november” and “december”?
Example What is the trigram overlap between “november” and “december”? novemberdecember nov ove vem emb mbe ber dec ece cem emb mbe ber
Example What is the trigram overlap between “november” and “december”? novemberdecember nov ove vem emb mbe ber dec ece cem emb mbe ber 3 trigrams of 6 overlap. How can we quantify this?
One option – Jaccard coefficient A commonly-used measure of overlap Let X and Y be two sets; then the J.C. is What does this mean? number of overlapping ngrams total n-grams between the two
Example novemberdecember nov ove vem emb mbe ber dec ece cem emb mbe ber 3 9 JC = 1/3
Jaccard coefficient Equals 1 when X and Y have the same elements and zero when they are disjoint X and Y don’t have to be of the same size Always assigns a number between 0 and 1 Threshold to decide if you have a match E.g., if J.C. > 0.8, declare a match
Efficiency We have all the n-grams for our query word How can we efficiently compute the words in our lexicon that have non-zero n-gram overlap with our query word? Index the words by n-grams! lo alonelord sloth
Matching trigrams Consider the query lord – we wish to identify words matching 2 of its 3 bigrams (lo, or, rd) lo or rd alonelordsloth lordmorbid bordercard border ardent Standard postings “merge” will enumerate … Adapt this to using Jaccard (or another) measure.
Context-sensitive spell correction Text: I flew from Heathrow to Narita. Consider the phrase query “flew form Heathrow” We’d like to respond: Did you mean “flew from Heathrow”? How might you do this?
Context-sensitive correction Similar to isolated correction, but incorporate surrounding context Retrieve dictionary terms close to each query term (e.g. isolated spelling correction) Try all possible resulting phrases with one word “fixed” at a time flew from heathrow fled form heathrow flea form heathrow Rank alternatives based on frequency in corpus Can we do this efficiently?
Another approach? What do you think the search engines actually do? Often a combined approach Generally, context-sensitive correction One overlooked resource so far…
Query logs How might we use query logs to assist in spelling correction? Find similar queries, e.g. “flew form heathrow” and “flew from heathrow” Query logs contain a temporal component! User will type “flew form heathrow” Not get any results (or any relevant results) User will issue another query “flew from heathrow” We can make this connection!
General issues in spell correction We enumerate multiple alternatives for “Did you mean?” Need to figure out which to present to the user Use heuristics The alternative hitting most docs Query log analysis + tweaking For especially popular, topical queries Spell-correction is computationally expensive Avoid running routinely on every query? Run only on queries that matched few docs
Dictionaries How do new ways to get information change our information needs? How does our reliance on Google for all different types of information affect it’s ability to provide specific useful information? How does the undermining of traditional sources of rules of English (and other languages) affect society’s use of English?